Position product

Position Product — Differentiation From Evidence

Produce ONE narrow, true, demonstrable claim about why this thing (or this person) over the alternatives — plus the proof plan that makes the claim land inside the audience's real attention window. Adjectives are not positioning. Evidence is.

When to use / when NOT to use

Use for: a product's core claim, a feature's "why this matters," a company one-liner, a portfolio thesis, a candidate's "why you" answer. The method is identical for products and people — a person's positioning is a product positioning where the artifact is their shipped work.

Do NOT use for:

• Competitor teardown / feature matrix → run-competitive-analysis (its output is step 3's input here)
• Narrating one piece of work end-to-end → tell-case-study-story (it inherits the thesis from here)
• Auditing whether EXISTING positioning claims hold up → craft-critique (this skill creates; that one judges)
• Designing the surface where positioning renders (landing page, onboarding, pricing) → design-growth-surfaces
• Final wording and register → writing-voice

Before you start (both mandatory)

1. Load craft-critique. Every claim about markets, competitors, buyers, recruiters, or trends in this method is handled by its evidence protocol (cite / get / flag) and screened against its banned constructions. Do not restate the protocol — apply it.
2. Load design-taste if the output will be designed. Its governing phrase — "quiet surface, loud proof" — is also the positioning law: restraint stays on the skin; the substance (real metrics, live artifacts, working systems) carries the persuasion.

Evidence tags (stamp one on every external claim)

Tag	Meaning	Allowed use in positioning
[V-high]	Survived adversarial verification, multiple independent sources	Load-bearing plank
[V-med]	Pattern verified; headline numbers vendor/marketing-grade	Supporting plank; never the thesis alone
[V-low]	Single small study / weak method	Design input only; flag if quoted
[GAP]	No claim survived checking	Never marketed; becomes a research task
[REFUTED]	Specifically disproven	Kill list — never appears, even paraphrased
[Assumption]	Chosen belief, revisable	Stated as an assumption, with what would change it

The method

1. Fix the audience and the moment

• Name the decider — recruiter, buyer, exec, first-time user. Not "users" in general.
• Establish the attention window with evidence, not hope. Worked case: a product designer positions their portfolio for a ~55-second recruiter first-pass that rarely includes an off-site click. Rule: position for the scan; assume the off-site click never happens.
• Name the real alternative they compare you against (incumbent, other candidates, doing nothing) and the default belief you must contrast — positioning is a correction to something they already think.
• Different audiences get different moments, not different truths. One fact base; per-audience emphasis.

2. Inventory only verified truth

List every fact the story might stand on. Grade each with a tag + source. Then apply three hard filters:

• Numbers without a named measurement source are presumed fabricated — cut, don't soften. Real case: plausible-sounding enterprise metrics ("25% satisfaction lift") in a candidate's old portfolio material were fabricated → cut entirely, not rounded down.
• Quotes and credentials must verify verbatim. Real case: a widely-repeated expert quote returned zero verbatim matches → the positioning re-anchored on the platform's official design guide instead. If the source can't be checked, swap the source, don't defend it.
• NDA'd / unshipped work keeps names + scale, drops outcomes. The "NDA-honest" pattern: client names and scope numbers ("7 clusters, 34 plants") stay; outcome metrics and "shipped to production" go unless verifiable.

Anything that fails a filter becomes either a GET (research it before continuing) or is out of the inventory. Positioning built on a fact you'd have to defend is already lost.

3. Map the field — supply gap × demand ask

Take run-competitive-analysis output, or run a scoped scan: what does every alternative claim, and what do they actually show?

• The highest-leverage seam is a convergence: something deciders explicitly ask for (demand-ask) that nobody in the studied field supplies (supply-gap). Worked case: hiring managers ask for outcomes in line one AND the studied competitor pool was weakest exactly there → "metric in line one" became the single highest-leverage move.
• Count the field. "0 of 61–73 studied portfolios host a live on-page AI demo" is a hard-evidenced scarcity claim. "Everyone is weak at this" is an impression — tag it honestly.

4. Find the seam — the candidate-claims table

Build this table before writing a word of copy:

Candidate claim	Evidence (tag + source)	Who else can truthfully say it	Copyable?

Copyability has two grades:

• Structural (best): the alternative's business model or incentives prevent copying. Worked case: an incumbent shopping assistant that blends sponsored results into conversation cannot become visibly fit-first without giving up a measured revenue lever — so "visibly-reasoned, fit-first" owns space the incumbent structurally can't enter.
• Scarcity-based (legitimate, decaying): nobody does it yet. Valid wedge; handle in step 6.

Pick the claim that is narrow (one seam, not general excellence), true (survives step 2), demonstrable (an artifact or number the audience can touch), and hardest to copy.

5. Write the thesis — one sentence, contrast form

The reliable shape: "Most X [the default the audience already believes]. This one [the difference] — and here it is running."

• Real person example: "Most product designers hand the AI off to engineers. I design it and I ship it in code — she's live on this page."
• Real product example: the shopkeeper who tells you the downside — an assistant that reasons from your stated needs, pins the runner-up, and refuses what it shouldn't sell.

Rules: no adjective does load-bearing work; the difference must be checkable within the moment from step 1; never name or disparage a competitor in the thesis — contrast the category default, generically and honestly.

6. Separate the scarcity fact from the winning bet (mandatory)

Every positioning output contains this split, explicitly:

• Hard-evidenced: the scarcity count, the measured metric, the structural incentive.
• Opinion/bet: that this difference will win. Almost always trend-supported opinion, not survey data. Say so.
• Decay: scarcity wedges lose value as they're copied → the recommendation is usually "ship while the field is empty," stated as a timing bet. Structural wedges hold while the incumbent's incentive holds — name the incentive so the expiry condition is visible.

Worked case, verbatim pattern: "the scarcity is hard-evidenced (the 0-of-61–73 scan); the 'this will win' is trend-supported opinion. Still the right bet — but a bet on timing, and its value decays as others copy it." That honesty flag IS the deliverable, not a hedge on it.

7. Plan the proof surface — quiet surface, loud proof

• Proof in line one. The metric, scale anchor, or live artifact leads; process and backstory follow. Never make the audience dig for the reason to believe.
• Demonstration beats description. A live, operable artifact collapses "prove the claim" into the scan itself; a link off-site is proof that mostly never gets clicked. If the claim is "we build X," the strongest surface is X, running, on the page.
• The no-title rule: minimal proof is a privilege of the already-credentialed — a frontier logo vouches for its people; if the audience doesn't already trust you, over-prove the substance while keeping the surface quiet (see design-taste for the surface register).
• Where no metric exists (NDA, early product): use the scale anchor and make the absence visible as discipline — an honest gap reads as integrity; a filled gap reads as fabrication the moment anyone checks.
• Never market a [GAP]/[REFUTED] claim, even the emotionally strongest one. Real case: an elder-companion product never markets loneliness outcomes — every study behind them was uncontrolled — so it's marketed as presence and connection, never clinical benefit.

8. Write the kill list

Enumerate what must never be claimed, with the reason and (where relevant) where the topic IS handled instead:

• [REFUTED] claims and fabricated/unsourced numbers
• Unverifiable third-party quotes or credentials
• Capabilities not yet shipped (re-add when shipped, not before)
• Topics deliberately kept off the face (e.g., visa/sponsorship handled in the application form, not the portfolio)
• Viral competitor stats with no method behind them — foil only, never a cited plank; find the better-grounded alternative and cite that instead

Positioning is as much refusal as assertion. Write the refusals down so future copy can't drift.

The differentiator test (all seven, or it's not the thesis)

• Narrow — claims one seam, not general excellence
• True — every load-bearing fact tagged [V-med] or better, with source
• Demonstrable — an artifact or number the audience can verify inside the attention window
• Scarce, evidenced — "0 of N studied" beats "few competitors"
• Hard to copy — structural if possible; if scarcity-only, decay stated
• Generic-honest contrast — no competitor named and disparaged
• Survives craft-critique's banned-constructions scan — zero "uniquely positioned" energy

Worked example — positioning a person (a product designer's portfolio thesis)

1. Audience & moment: recruiter/hiring manager · ~55s reviewer first-pass, rarely clicks off-site (the portfolio's working assumption) · compared against AI-role designers at frontier AI companies · default belief: "designers describe AI work; engineers build it."
2. Truth inventory: shipped, live AI chatbot with real measured conversion metrics [V-high — first-party analytics, sources named] · enterprise work under NDA — client names + scale kept, outcome metrics cut as unverifiable, "shipped to production" claim FALSE → cut · a graduate degree completing on a verified date · plausible-sounding old metrics found fabricated → deleted, not softened.
3. Field map: a competitive scan of 61–73 AI-role designer portfolios (incl. frontier-company designers): 0 host a live on-page AI demo — AI proof is always described or linked off-site [hard-evidenced scarcity]. Converges with the demand-ask: reviewers want outcomes in line one.
4. Seam: "designs AND ships the AI" — narrow, true (live artifact), demonstrable within the scan (the demo runs on the page), and hard to copy quickly (requires actually building).
5. Thesis: "Most product designers hand the AI off to engineers. I design it and I ship it in code — and you're standing in the proof of it."
6. Scarcity vs bet: scarcity hard-evidenced (the 0-of-N scan); "a live demo wins the role" is trend-supported opinion; value decays as copied → ship while the field is empty, named as a timing bet.
7. Proof plan: flagship case leads with the real measured metric in line one · NDA case leads with the scale anchor, no numbers, absence-as-discipline · the live operable demo is the wedge, on-page, so proof lands inside the 55 seconds.
8. Kill list: fabricated enterprise metrics (cut) · "uniquely positioned" constructions (banned) · unshipped skills (off until shipped) · sponsorship off the portfolio face (handled in the application form) · nonexistent projects (never referenced).

Worked example — positioning a product (compressed)

Step	Honest conversational-commerce assistant (Echo Show, first-time internet users)
Audience & moment	Low-literacy first-time buyers who trust peers, not institutions [V-high] — and the exec approving the product
Truth	Honest by architecture: zero invented prices by construction; medicine questions refused by construction (verifiable in the build, not a promise)
Field	Incumbent assistant blends sponsored results into conversation; ranking opaque to shopper and merchant [V-high]; personalization is a measured revenue lever pulling every incumbent sell-first [V-high]
Seam	Visibly-reasoned, fit-first recommendation — structural: the incumbent can't copy it without giving up the revenue lever. "The credibility no blended-ranking assistant can copy."
Thesis	The shopkeeper who tells you the downside: reasons from your stated needs, pins the runner-up, refuses what it shouldn't sell
Scarcity vs bet	Honesty-by-architecture is verifiable now; "trust converts this segment" rests on peer-influence data [V-med] + a bet — flagged
Kill list	No loneliness-outcome claims ([REFUTED] marketing — every study uncontrolled) · the viral "incumbent is 83% self-serving" stat = single vendor quote, no method → narrative foil only; cite the grounded 1,300-pick analysis instead

Anti-patterns

Don't	Why it fails	Do
"Uniquely positioned…" / "strong foundation…"	Banned per craft-critique — fits anyone	One narrow checkable claim
Adjective positioning ("seamless," "best-in-class")	Unfalsifiable = unpersuasive	Lead with the artifact or number
Filling an NDA/metric hole with a plausible number	Fabrication detected = everything else discounted	Scale anchor + visible absence as discipline
Resting the story on an unverified quote/credential	Collapses when checked	Verify verbatim or swap to the official citable source
Naming and trashing the competitor	Reads insecure; legal/aggro risk	Contrast the category default, generically
Leading with process/backstory	The scan ends before the proof arrives	Proof in line one
Treating "this will win" as evidenced because scarcity is	Conflates fact with bet	Split them; state the decay
Marketing the [REFUTED]-but-moving claim	One checked claim poisons all claims	Kill list, enforced
One positioning blob for every audience	Nobody's moment is served	One fact base, per-audience emphasis (step 1)
Differentiation buried in a feature list	Ten claims = no claim	One seam, one sentence

Output format

## Positioning: [product / person / feature]

**Audience & moment:** [decider · attention window (evidence) · compared against · default belief]

**Thesis:** [one contrast sentence — "Most X… This one… — and here it is running."]

**Claims table:**
| Candidate claim | Evidence (tag + source) | Who else can say it | Copyable? |

**Scarcity vs bet:** hard-evidenced: […] · opinion/bet: […] · decay/expiry condition: […]

**Proof plan:** [line-one proof per surface; live demonstration > description; absence-as-discipline where no metric exists]

**Kill list:** [claims that never appear + why + where the topic is handled instead]

**Under-evidenced flags:** [every `[GAP]`/`[Assumption]` + exactly what data would strengthen each]

Sources

• Method, evidence tags, kill-list discipline — Dinesh's own positioning practice; not a fork, not NN/g-grounded. The worked examples are grounded in real practice: the live-AI-demo wedge and the 0-of-61–73 portfolio scan (a product-designer portfolio strategy); the NDA-honest scale-anchor pattern (an enterprise-UX resume spine — client names + scale kept, outcomes and "shipped to production" cut); the honest-by-architecture conversational-commerce assistant with its foil-vs-cite split.
• craft-critique — single source of truth for the evidence-discipline protocol (cite / get / flag) and the banned constructions; this skill applies both and never restates them. The [V-*]/[GAP]/[REFUTED] tags here are a positioning-specific grading layer on top of that verification, not a copy of it.
• design-taste — owner of "quiet surface, loud proof," reused here as the positioning law by reference.

Boundaries

• craft-critique owns the evidence protocol and banned constructions — this skill applies them, never restates them.
• run-competitive-analysis owns the competitor teardown; its output feeds step 3.
• tell-case-study-story owns narrating a single piece of work; it inherits this skill's thesis and proof plan.
• design-growth-surfaces owns the surfaces (landing, onboarding, pricing) where positioning renders.
• design-taste owns the surface register, including "quiet surface, loud proof."
• writing-voice owns the final wording of any outward copy.